@ragpipe/plugin-ollama

@ragpipe/plugin-ollama

Ollama local embedding and generation plugin for ragpipe.

No API key required. Pair with @ragpipe/plugin-sqlite-vec for a fully local RAG setup.

Prerequisites

# Install and start Ollama
ollama serve

# Pull models
ollama pull bge-m3
ollama pull llama3

Install

pnpm add ragpipe @ragpipe/plugin-ollama

Usage

import { defineConfig } from "ragpipe";
import { ollamaEmbedding, ollamaGeneration } from "@ragpipe/plugin-ollama";

export default defineConfig({
  embedding: ollamaEmbedding({
    model: "bge-m3",
    dimensions: 1024,
  }),
  generation: ollamaGeneration({
    model: "llama3",
  }),
  // ... vectorStore
});

Remote Ollama server

ollamaEmbedding({
  model: "bge-m3",
  dimensions: 1024,
  baseUrl: "http://gpu-server.local:11434",
});

API

ollamaEmbedding(options)

Returns an EmbeddingPlugin that calls Ollama's /api/embed endpoint.

Option	Type	Default	Description
model	string	—	Embedding model name (required)
dimensions	number	—	Vector dimensions (required)
baseUrl	string	"http://localhost:11434"	Ollama server URL

• Batch support: embedMany() sends an array of texts in a single API call

ollamaGeneration(options)

Returns a GenerationPlugin that calls Ollama's /api/chat endpoint.

Option	Type	Default	Description
model	string	—	Generation model name (required)
baseUrl	string	"http://localhost:11434"	Ollama server URL
systemPrompt	string	"Answer based on the provided context."	Default system instruction

• Streaming: generateStream() returns an AsyncIterable<string> via NDJSON
• History: Pass { history } to include conversation context
• Per-call override: Pass { systemPrompt } at call time to override the default

Supported models

Embedding

Model	Dimensions
bge-m3	1024
nomic-embed-text	768
mxbai-embed-large	1024
all-minilm	384

Generation

llama3, llama3.1, mistral, gemma2, phi3, qwen2, and any model available via ollama pull.

License

MIT